Anthropic launches Claude Fable 5: the public version of its most powerful model with safety guardrails

Anthropic launched Claude Fable 5, the first public version of its Mythos-class model — a category of models that sits above the Opus line in capability. It's the same model as Mythos 5, but with specific guardrails that allow its release to the general public.

Why does it have guardrails if it's Mythos?

Anthropic unveiled Mythos 5 in April 2026 and limited its rollout to government and industry partners due to its advanced cybersecurity capabilities. The company even issued a public warning that AI was becoming "too dangerous" without adequate controls. Fable 5 resolves that tension: it uses the same underlying technology but with specific blocks in high-risk areas — cybersecurity, biology and chemistry — which allows it to be released without the restrictions imposed on Mythos.

What can it do?

Fable 5 excels at software engineering, knowledge work, and visual analysis. In benchmarks, it outperforms Claude Opus 4.7 on the most complex coding tasks. The difference is that where Mythos might answer questions about finding vulnerabilities in real systems, Fable blocks that type of query.

Price and availability

Via API: $10 per million input tokens and $50 per million output tokens. For subscribers (Pro, Max, Team, Business, Enterprise), access is enabled progressively through June 22. The model is already available on the Claude API.

Context

Fable 5's release comes weeks after Anthropic warned about the risks of AI models that are too powerful without controls. It also comes as the company prepares for its IPO. The strategy of launching a "safe" version of its most advanced model is both a safety move and a commercial one: it democratizes access to frontier technology while keeping the full model (Mythos) as an offering for select partners.